Growth Inhibition of Sodium Chloride Crystals by Anticaking Agents: In

Nov 19, 2012 - two anticaking agents for sodium chloride powders, namely, ferrocyanide [K4Fe(CN)6] and iron(III) meso-tartaric acid (Fe-. mTA; 1:1 mol...
0 downloads 0 Views 1MB Size
Download PDF
Article pubs.acs.org/crystal

Growth Inhibition of Sodium Chloride Crystals by Anticaking Agents: In Situ Observation of Step Pinning Arno A. C. Bode,*,† Shanfeng Jiang,‡ Jan A. M. Meijer,‡ Willem J. P. van Enckevort,† and Elias Vlieg† †

Radboud University Nijmegen, Institute for Molecules and Materials, 6500 GL Nijmegen, The Netherlands Akzo Nobel Industrial Chemicals, Salt and Crystallization, 7418 AJ Deventer, The Netherlands

‡

S Supporting Information *

ABSTRACT: We have determined how the growth of sodium chloride crystals is inhibited by the anticaking agents ferrocyanide and iron(III) meso-tartaric acid. Using in situ atomic force microscopy, we show how steps flow on clean crystals and how the step flow is inhibited by the anticaking agents. At submonolayer coverages, steps are temporarily pinned, whereas at higher coverages, no step movement is observed at all. Finally, we found that the two anticaking agents influence the surface morphologies of treated crystals differently, providing a possible explanation for the observation that powders treated with ferrocyanide flow freely whereas those treated with iron meso-tartaric acid show slip−stick behavior.

■

INTRODUCTION

rate, morphology, and nucleation rate of sodium chloride crystals was already studied in 1965.10 Ferrocyanide is by far the most frequently used anticaking agent for sodium chloride in Europe. It has some drawbacks, however. Most of the sodium chloride produced is used for the production of chlorine gas by electrolysis. Because of their stability and ionic nature, ferrocyanide ions are difficult to remove prior to electrolysis. During electrolysis, the iron causes the formation of iron hydroxyde [Fe(OH)3] on the electrodes and at the membranes, increasing power consumption. Furthermore, because ferrocyanide contains nitrogen, it is a source of nitrogen trichloride, an explosive gas. Therefore, a new anticaking agent for sodium chloride was introduced: Fe-mTA.11 This anticaking agent is nearly as effective as ferrocyanide; however, the complex is less stable, so the iron can more easily be removed from the brine prior to electrolysis by adding lye. This prevents corrosion of the electrodes and membranes and significantly reduces power consumption. Furthermore, meso-tartaric acid does not contain nitrogen, and therefore, less nitrogen trichloride gas is formed. Recently, we showed how the anticaking agent ferrocyanide firmly adsorbs on the sodium chloride {100} surface using surface X-ray diffraction.9 In that study, it was concluded that the ferrocyanide ions replace sodium chloride clusters on the surface and block further growth of the NaCl crystal because of the difference in ionic charge. In this way, caking is prevented. The required amount of ferrocyanide is 1.4 × 10−6 mol per

Crystal growth inhibitors are subject to intensive study because they are important in many applications. They provide control over crystal shape and size distributions, such as preventing the formation of needle-shaped crystals.1,2 They can also be used to prevent agglomeration of crystalline powders (i.e., caking).3 In this case, the inhibitors are called anticaking agents.4 However, their functioning is not well understood. In essence, two basic theories have been proposed, both assuming adsorption of the inhibitor molecules on the crystal surface: Either individual molecules adsorb and block growth by step movement on the crystal surface,5 or the whole surface is covered by a “protecting” adsorption layer.6 Only in a few cases is it known how a certain growth inhibitor adsorbs onto the crystal surface.7−9 Here, we investigate the growth inhibition of two anticaking agents for sodium chloride powders, namely, ferrocyanide [K4Fe(CN)6] and iron(III) meso-tartaric acid (FemTA; 1:1 molar mixture of FeCl3 and meso-C4O6H6 at pH 4− 5), which are known to be very effective at low concentrations.10,11 Sodium chloride (NaCl) is a very important base compound for the chemical industry and is therefore produced in very large amounts. However, sodium chloride crystals easily cake during storage and transportation.3 As this makes handling of the material very difficult, caking must be prevented through application of an anticaking agent. For sodium chloride, many anticaking agents are known, including sodium and potassium ferrocyanides, nitrilotriacetamide [N(CH2CONH2)3], cadmium chloride (CdCl2), and sodium metaphosphate [(NaPO3)n, n ≈ 16]. The influence of these anticaking agents on the growth © 2012 American Chemical Society

Received: May 25, 2012 Revised: October 16, 2012 Published: November 19, 2012 5889

dx.doi.org/10.1021/cg3012537 | Cryst. Growth Des. 2012, 12, 5889−5896

Crystal Growth & Design

Article

mole of sodium chloride, which corresponds to about 2 × 10−6 mol/m2, assuming an average particle size of 200 μm. This amount is almost identical to the maximum coverage of ferrocyanide ions on the (100) surface of sodium chloride,9 which corresponds to 2.6 × 10−6 mol/m2. This we define as a coverage fraction of θ = 1. The required amount for Fe-mTA is in the same range as for ferrocyanide, about 3 × 10−6 mol per mole of NaCl. For Fe-mTA, the maximum coverage has not been determined, but because the size of the Fe-mTA complex is similar to that of ferrocyanide, a monolayer will be of the same order of magnitude. Therefore, also for Fe-mTA, we define the coverage fraction θ = 1 as 2.6 × 10−6 mol/m2. A low-energy ion spectroscopy (LEIS) study indicated the presence of iron oxochloride and iron oxide layers and a layer of organic material on sodium chloride crystals treated with FemTA.12 In that study, performed in a vacuum, it was assumed that the iron atoms are transported toward the sodium chloride surface by the Fe-mTA complex, where they form a layer of iron oxochloride and a layer of iron oxide on top. These layers would inhibit the growth of the sodium chloride and thus prevent caking. In this model, it was assumed that the only role of the meso-tartaric acid molecules is to prevent the formation of iron oxide in solution, by forming an organometallic complex with the iron. Here, we investigate the growth of sodium chloride crystals by studying step movement on the crystal surface using atomic force microscopy (AFM). Normally, such studies are only possible using a liquid cell.13−17 However, in 1996, Shindo et al.18,19 studied the adsorption of water on the sodium chloride {100} surface using AFM. They showed that, at a relative humidity (RH) of 52% or higher, monatomic steps become mobile. Thereby, they showed that, at this humidity, well below the deliquescence point of 75% RH, the adsorbed water layer is thick enough to allow solvation and transport of ions across the surface, even though the water layer is estimated to be only 1−4 nm thick.20−22 Because the water layer is very thin, the AFM tip is not completely submerged. In this work, we used this effect to study the growth of sodium chloride crystals and how it is influenced by the two anticaking agents by in situ AFM. Therefore, no liquid cell was needed in these experiments. Step movement occurred by ion transportation through the adsorbed water layer, as the RH was kept at 58%. This condition is also close to that applied for the industrial usage of anticaking agents. Apart from the in situ experiments, we also determined how the anticaking agents change the morphology of the crystal surface at the industrially applied concentration, because, in practice, it was found that the two anticaking agents change the flow characteristics of sodium chloride powders differently. Flow properties are usually influenced by the surface morphology of a crystalline powder.23 Even though powders treated with both agents do not cake, a powder treated with ferrocyanide flows freely, whereas a powder treated with FemTA shows slip−stick behavior.

■

Step Flow Experiments. A Digital Instruments Nanoscope IIIa MultiMode AFM instrument was operated in contact mode using a Jtype piezo scanner. Commercial silicon nitride contact-mode cantilevers were used. Both height and deflection signals were recorded. Even though measurements were performed in contact mode, the surface was not altered by the imaging. After the measurements, no scan effects were observed, which was checked by making a 50 × 50 μm2 scan around the imaged area. The microscope was placed in a plexiglass box to control the humidity. A dish of saturated sodium bromide solution in water was placed in the box to keep the RH at 58 ± 3%, which was monitored with an EL-USB-2 data logger. The temperature in the box was 25 ± 1 °C. Variation of the temperature will have little influence on the crystal growth, because the solubility of sodium chloride is nearly independent of temperature.3 A large-area scan, 50 × 50 μm2, was performed to select a sloping area containing many monatomic steps. Zooming in on this area, 4 × 4 μm2 images were captured continuously at the same location, creating a film. Each image took 512 s to capture (512 lines, 1 Hz), and only images in the same scanning direction were used for the film. Scanning was continued for 1−2 days, giving 80−160 frames per film. The result is a sped-up film of step flow, at 1024 s per frame. The anticaking agents were applied using small droplets of volatile solvents (ethanol or methanol) in which sodium chloride is poorly soluble, to minimize surface roughening by dissolution of the sodium chloride crystal. After evaporation of the solvent, the sample was placed in the humid environment. To check the influence of solvent, films of moving steps on the surfaces were recorded after evaporation of a 2-μL droplet of clean ethanol or clean methanol without anticaking agents, applied using a Gilson P20 micropipet. Even though the surfaces treated with methanol showed slightly more roughening than those treated with ethanol, no significant difference in the rate of step movement was observed between these preparations. Having characterized the properties of the clean system, films of steps and step flow were obtained for several concentrations of applied anticaking agents. For Fe-mTA, coverages in the range θ = 0.005−46 were used. Ethanol was used as the solvent for Fe-mTA because the solubility of sodium chloride in ethanol is very low, leading to minimum surface roughening during the evaporation of the solvent droplet. Surface roughening was further limited by using droplets of only 2 μL. For ferrocyanide, the applied coverages ranged from θ = (8 ± 4) × 10−5 to θ = 17. The lowest concentrations of ferrocyanide were applied using ethanol as the solvent. Because of the extremely low solubility of potassium ferrocyanide in ethanol, methanol was used as a solvent for the high concentrations of ferrocyanide. This caused slightly more surface roughening but did not change the step flow rate. Surface roughening was again limited by applying 2-μL droplets. The industrially applied amounts of anticaking agent on chemical transformation salt correspond to about 2 × 10−6 mol/m2 for both ferrocyanide and Fe-mTA, which is approximately one monolayer. The high end of the concentration ranges investigated thus corresponds to very large amounts of the anticaking agent, much more than required in industry for effective anticaking. The lowest concentrations used were several orders of magnitude lower, chosen such that any difference with clean crystals could no be longer observed. Ex Situ Experiments. Cleaved sodium chloride {100} surfaces treated with the anticaking agents were examined in more detail ex situ, to study the surface morphology caused by the anticaking agents. The applied coverages of the anticaking agents were θ = 0.8, 0.4, and 0.04. The anticaking agents were applied using a 2-μL droplet of methanol. Methanol was used for both anticaking agents to eliminate differences in initial surface roughening by the different solvents. The methanol was allowed to evaporate, and then the crystals were exposed to 58% RH for a few days. After the crystals had been removed from the plexiglass box, the resulting surface morphology was examined, ex situ, by AFM at a humidity of approximately 40%. In parallel with these experiments, crystals were treated with the same amount of anticaking agents and examined by AFM immediately after evaporation of the methanol, to verify that the morphology change did not result from recrystallization in the methanol droplet.

EXPERIMENTAL SECTION

Specimen Preparation. Sodium chloride {100} surfaces (approximately 3 × 3 mm2) were obtained by cleaving a melt-grown sodium chloride crystal, purchased from Ted Pella, Inc. (Redding, CA, USA). The cubic {100} faces are the only relevant faces of sodium chloride crystals for anticaking, because the anticaking agents are applied on crystals after their growth. Such crystals have a cubic morphology for sodium chloride. 5890

dx.doi.org/10.1021/cg3012537 | Cryst. Growth Des. 2012, 12, 5889−5896

Crystal Growth & Design

Article

These ex situ observations were performed using a Digital Instruments Dimension 3100 AFM instrument operated in both tapping and contact modes. The resulting surface topography was studied in more detail by acquiring images at various scales, ranging from 100 × 100 nm2 to 100 × 100 μm2.

0.281 nm. These observations are in agreement with the results of Shindo et al.19 In these experiments, we found that the tops of hillocks were dissolving, whereas the steps near the foot of the same hillock were growing. Therefore, the main driving force for step movement in these experiments is the minimization of step energy by the elimination of height differences: The surface tends to flatten. This process, surface relaxation,24 has been reported for metal and semiconductor surfaces,25−27 as well as for boric acid.28,29 It is closely related to Ostwald ripening,30 which is the minimization of surface energy, or in this case, step energy. We selected locations at the foot of hillocks to have the same conditions in all experiments. If surface relaxation is the main driving force for step movement, the overall supersaturation (Δμ) should be close to zero. Therefore, we estimated the overall supersaturation of the solution nanolayer on the clean surface at a humidity of 58%, by measuring the rate of step movement and the radius of curvature of freely flowing steps on the clean crystal surface. The rate of step movement was found to be independent of the terrace width in all experiments; thus, the growth rate is reaction-limited. Therefore, according to the Burton−Cabrera− Frank (BCF) theory of crystal growth,31 the rate of step movement and local radius of curvature of a step are related by

■

RESULTS AND DISCUSSION To determine whether the anticaking agents are incorporated into the crystal lattice or desorbed from the surface during growth, NaCl crystals were grown from aqueous solutions in the presence of from 6.1 × 10−5 to 6.1 × 10−2 mol/L (1−1000 ppm) ferrocyanide and Fe-mTA. The resulting crystals were removed from the growth solution, dried, and subsequently dissolved in water. The iron contents of these solutions were determined using inductively coupled plasma mass spectrometry (ICP−MS), along with collision cell technology (CCT). Only a very small amount of iron could be detected in both cases, corresponding to the iron present in the adhesive water layer. Therefore, we conclude that almost no iron is incorporated. The anticaking agents are reversibly adsorbed onto the surface and desorb from the surface when growth continues. Step Movement on a Clean Surface. A film was acquired showing the propagation of monatomic steps on a clean sodium chloride crystal surface at RH = 58%. Also, when the crystal surface was pretreated with a 2-μL droplet of either ethanol or methanol, monatomic step movement was observed. This result is shown in Figure 1 for a surface treated with a droplet of methanol. Several films of step propagation on clean crystal surfaces were recorded, where the surfaces were pretreated with either methanol or ethanol. Two complete films of step flow on clean crystals are provided as Supporting Information. The measured step height was approximately 0.3 nm, which corresponds to the monatomic step height of 1/2d⟨100⟩ =

⎛ ρcrit ⎞ ⎟⎟ vstep( r ⃗) = bΔμeff ( r ⃗) = bΔμ⎜⎜1 − ρcurve ( r ⃗) ⎠ ⎝

(1)

where vstep is the local rate of step movement at location r;⃗ b is a kinetic constant; and Δμeff is the local effective supersaturation, which differs from the overall supersaturation Δμ because of the local curvature. ρcurve is the local radius of curvature of the step, and ρcrit is the critical radius for two-dimensional nucleation on the surface, given by ρcrit =

γstepΩ hstepΔμ

(2)

whereas Ω is the volume of a growth unit [(1/2d⟨100⟩)3 = 0.0222 nm3]; hstep is the step height (0.281 nm); and γstep is the step free energy, which is approximately 1.1 × 10−20 J/nm for monatomic steps on the {100} face of sodium chloride crystals in contact with an aqueous solution.32 The propagation rate of the freely moving steps in Figure 1, as indicated by the arrows, was determined together with their radius of curvature. In Figure 2, the rate of step movement of these steps is plotted against the radius of curvature, and eqs 1 and 2 are fitted to the data. The best fit gives a supersaturation of Δμ = (1.6 ± 0.6) × 10−24 J, or Δμ/(kT) = (4 ± 1) × 10−4, where k is the Boltzmann constant and T is the absolute temperature. This supersaturation is very low, in agreement with the observation that the main driving force for step movement is surface relaxation. The critical radius ρcrit is 550 ± 200 nm. Because this radius is very large, we can exclude twodimensional nucleation under these conditions, in agreement with our observations. The kinetic constant b was found to be (1.3 ± 0.6) × 1023 nm/(J s). The overall supersaturation Δμ is probably induced by local evaporation of the water layer due to heating caused by the laser light. Another contribution to the supersaturation is the capillary action of the tip. The capillary action increases the local water layer thickness around the tip. When the tip has moved away, the dissolved ions in the attracted water will

Figure 1. Step movement on a clean crystal surface treated with a clean droplet of methanol, 4 × 4 μm2 AFM height images. (a) Step is pinned in lower part of image. (b−d) As soon as the pinned point is passed, the highly curved step quickly moves to reduce its negative curvature. Other monatomic steps also move, but at a much lower rate. 5891

dx.doi.org/10.1021/cg3012537 | Cryst. Growth Des. 2012, 12, 5889−5896

Crystal Growth & Design

Article

expected because the lowest-energy Burgers vector in the sodium chloride F cell is 1/2⟨110⟩. Growth Inhibition by Anticaking Agents. Ferrocyanide. Having established the method of observing step movement, we tested how anticaking agents hamper or block step propagation. At high coverages of ferrocyanide, above θ = 3.5, the surfaces were found to be very rough, and no steps were visible. Figure 4 shows a typical surface morphology

Figure 2. Propagation rate of the freely moving steps in Figure 1 plotted against the local radius of curvature. The continuous line is the best fit of eqs 1 and 2. The data points are single measurements; the error is represented by the dot size.

contribute to the local supersaturation, because the layer thickness will return to its equillibrium value.18 Even on the clean surface, several steps are pinned by dislocations and other obstacles, because of the extremely large critical radius of curvature of 500 nm.19 Such steps are shown in the left part of the images in Figure 1. We determined the average rate of step propagation and the local radius of curvature of such steps in several films of step propagation on the clean surfaces to compare the supersaturation to the value obtained for the freely moving step. Even though the steps were pinned, the step movement was not zero between pinning points (Figure 3). Overall, the step rate was 0.020 ± 0.005 nm/

Figure 4. Sodium chloride surface roughened by the application of a high coverage of Fe-mTA (θ = 47). The surface is very rough, and no steps are visible. The surface does not change when exposed to 58% RH. Surfaces roughened by a large amount of ferrocyanide look identical.

(obtained using Fe-mTA). A sample treated with ferrocyanide (θ = 0.7, slightly less than the industrially applied amount) showed less roughening and some steps that did not move at all at 58% RH. At a lower coverage of θ = 0.35, no roughening was observed. Only monatomic steps were seen on the surface, but they were fully blocked by the ferrocyanide ions. Only at a much lower coverage of θ = 2 × 10−4 was step movement observed, together with their temporary pinning due to the presence of the ferrocyanide ions. This is shown in Figure 5; the full film is available in the Supporting Information. The step indicated by the arrows in Figure 5 is negatively curved, as the step is pinned and its movement is stopped for about 1 h. Nevertheless, the net step movement was not retarded at this coverage of ferrocyanide because the highly curved step rapidly caught up with the other steps when the pinning molecule was desorbed. The net propagation rate of this step was approximately 0.02 nm/s. Calculated from the coverage and assuming complete adsorption, only 5 × 103 ferrocyanide ions were present on the imaged area, resulting in an average spacing of about 90 nm. Therefore, pinning of steps probably occurs by individual or small clusters of ferrocyanide ions. Extrapolating the observation of step pinning at θ = 2 × 10−4 to higher ferrocyanide coverages clearly shows that complete step blocking occurs at such coverages.

Figure 3. Step rate and local radius of curvature of a step pinned at two points. Even though the step is pinned, the rate of step movement between the pinning points is not zero.

s. According to eq 1, this rate corresponds to an average radius of curvature of 550 ± 200 nm. This agrees well with the observed radius of curvature of 600−800 nm for these pinned steps. No significant difference was found between ethanol- and methanol-treated samples. Many steps emerged from screw dislocations. As their height is 1/2d⟨100⟩, they must have a Burgers vector (b⃗) component perpendicular to the (100) face of 1/2[100]. This is to be 5892

dx.doi.org/10.1021/cg3012537 | Cryst. Growth Des. 2012, 12, 5889−5896

Crystal Growth & Design

Article

Figure 5. Step pinning by ferrocyanide (θ = 2 × 10−4). Images are 4 × 4 μm2 AFM height images. In image a, the step indicated by the arrow is pinned. In image b, the pinning point is passed, and the step pins at another point. This sequence repeats in images c and d. The intervals between images are 85, 50, and 20 min respectively.

Figure 6. Height image (4 × 4 μm2) of steps on a sodium chloride crystal treated Fe-mTA, θ = 1. Steps are several atomic layers high and are highly curved. They are immobile when exposed to 58% RH air.

When a second step passed the point where a previous step was pinned, it was most often not pinned at the same location, showing that the pinning point was removed. Therefore, we conclude that the pinning was caused by molecules that desorbed and that the pinning of the step was not caused by an edge dislocation. At an even lower coverage of θ = 8 × 10−5, no significant differences from the clean crystals were detected. Fe-mTA. A similar series of coverages was used to examine the influence of Fe-mTA on the growth of sodium chloride crystals. At high coverages, θ = 47 and θ = 5, a very thick, rough layer was observed on top of the crystal, as shown in Figure 4. No steps were visible at all, nor was there any change in the surface during exposure to humid air (58% RH). At a lower coverage of θ = 1, approximately corresponding to the concentration used industrially, steps were visible on the surface. However, they were several atomic layers high and much less smooth than those observed on clean crystals; see Figure 6. Furthermore, they did not move at all upon exposure to humid air (58% RH). The steps were highly curved and bunched, with small, round holes in the layers. Such a morphology is typical for surfaces blocked by step pinning.14 At lower coverages, monatomic steps were present, which were mobile. At θ = 0.5, movement of these monatomic steps was observed. They were, however, pinned at many locations by the adsorbed Fe-mTA. An example is shown in Figure 7; the full film is available in the Supporting Information. Even though most pinning points were spiral dislocations, some steps were pinned at sites without a dislocation outcrop. The arrows in Figure 7a,b indicate a location where a step was negatively curved but was temporarily not moving because of pinning. In Figure 7c,d, the pinning point was removed, and the step moved again. Similarly to the case of low coverage of ferrocyanide, no net retardation of the step movement was observed. This step moved rapidly after the desorption of the

Figure 7. Step pinning by Fe-mTA: (a) 4 × 4 μm2 height image of a sodium chloride surface treated with Fe-mTA, θ = 0.5. Steps are pinned at many locations, most are dislocations. (b−d) Enlargements of the indicated 1 × 1 μm2 area in image a: The step indicated by the arrow is pinned in image b and moving again in images c and d. Intervals between images b, c, and d are 34 min.

blocking molecules, catching up with the other steps. The net step rate was 0.028−0.056 nm/s, which is somewhat higher than expected. This is possibly explained by the relative surface steepness of the location, creating a supersaturation that was locally higher than that on flatter areas. 5893

dx.doi.org/10.1021/cg3012537 | Cryst. Growth Des. 2012, 12, 5889−5896

Crystal Growth & Design

Article

of Figure 8a. Both the dendrite patterns and the hillocks were nearly square, with the edges aligned parallel to the crystal lattice axes. Crystal surfaces treated with Fe-mTA were influenced very differently. Figure 8c shows the surface morphology induced by Fe-mTA, θ = 0.8. This surface was very rough, with high steps (up to 30 nm) and hillocks up to 15 nm high and 500−800 nm wide. In contrast to the surfaces treated with ferrocyanide, these features were rounded. At a lower coverage of Fe-mTA, θ = 0.4, similar rounded features were observed (Figure 8d). However, these hillocks were smaller, 2−5 nm high and 200−500 nm wide, and more numerous. This indicates that, at a lower concentration, nucleation of sodium chloride was blocked less effectively, resulting in more and smaller surface features. The observed features were formed during drying after exposure to the humid atmosphere, because these features were not observed when the crystals were studied immediately after application of the anticaking agents. The volume of the observed features was quite high compared to the solubility of sodium chloride and the estimated amount of adsorbed water. However, the thickness of the water layer on sodium chloride crystals at 58% RH is not well-known. Estimations range from 3 to 10 water layers,20−22 about 1−4 nm thick. Also, the solubility of sodium chloride in such a layer can differ from a bulk water solution. So, these features can indeed be formed during evaporation of the saturated adsorbed water layer. The surface morphology is very different for surfaces treated with ferrocyanide compared to those treated with Fe-mTA. Surface morphology is very important for powder flow characteristics; therefore, this is probably the reason why salt powders treated with ferrocyanide have different flowability properties than powders treated with Fe-mTA. Comparing Blocking Effects of the Anticaking Agents. The observed growth inhibition by the anticaking agents ferrocyanide and Fe-mTA is very similar, even though the iron complexes are very different. Both anticaking agents are able to pin monatomic steps on the surface of sodium chloride. At concentrations corresponding to the amounts used industrially for optimal anticaking effectiveness, both anticaking agents fully block step movement. These concentrations correspond to a (partial) monolayer of adsorbed iron complexes. Both anticaking agents can also retard steps. Ferrocyanide ions are able to pin steps at very low coverage (θ = 2 × 10−4). According to the step pinning model,5 step flow is blocked if the average distance between adsorbed pinning molecules is less than twice the critical radius of curvature. Therefore, the effective concentration of a blocking additive will be highly dependent on the supersaturation. In industrial applications, a much higher concentration of ferrocyanide is required. This is probably because of the much higher supersaturations reached during storage, because the critical radius decreases with increasing supersaturation (see eq 2). Because the critical radius of curvature at high supersaturations is very small for sodium chloride, the distance between adsorbed molecules has to be smaller to pin the steps. In the limit in which the critical radius approaches the size of the pinning agent, the step pinning mechanism is the same as the insulating adsorption layer mechanism. The required concentration is increased because adsorbed impurities can be desorbed from the surface. Even though the original theory on step pinning did not include impurity desorption,5 the theory has been adapted to include this effect.14,33

When the coverage of Fe-mTA was lowered further, step pinning was no longer observed. At coverages of θ = 0.005 and 0.05, no difference from the clean crystals could be observed. Ex Situ Experiments. To examine the crystal surfaces after evaporation of the adsorbed water layers, ex situ AFM was applied. When the crystals were examined directly after application of the anticaking agent, only a little roughening of the surface was detected. After exposure to 58% RH and drying, the surface morphology was dramatically different when anticaking agents were present. If no anticaking agent was applied, the surface remained smooth. Application of high amounts of anticaking agents (i.e., coverages of θ = 1 or higher) resulted in rough surfaces, such as the surface shown in Figure 4. The surface morphologies resulting from lower coverages are shown in Figure 8. Some more images at different magnifications are given in the Supporting Information.

Figure 8. Surface morphology of sodium chloride crystals treated with anticaking agents and exposed to 58% RH. (a) Ferrocyanide, θ = 0.4: dendritic growth in the ⟨110⟩ directions from one nucleation site. (b) Ferrocyanide, θ = 0.04: nucleation at multiple sites; inhibition less effective than in a. (c) Fe-mTA, θ = 0.8: rough layer of rounded islands of sodium chloride formed. (d) Fe-mTA, θ = 0.4: many rounded hillocks but less rough than in c.

Figure 8a shows the surface morphology induced by ferrocyanide at a coverage of θ = 0.4. Crystal growth on the surface was blocked by the ferrocyanide ions adsorbed onto it; therefore, dendritic growth of sodium chloride on top of the surface took place, starting from one nucleation site and spreading across the surface in the ⟨110⟩ directions. This is comparable to the ⟨111⟩ dendritic growth of sodium chloride crystals in the presence of ferrocyanide.10 These dendrites were composed of square, centrally depressed patterns, 20−40 nm high and 200−900 nm wide. In Figure 8b, the surface is shown as formed after addition of ferrocyanide at a coverage of θ = 0.04. Several isolated hillocks can be observed, so nucleation at the surface was probably not inhibited as effectively as in Figure 8a. These hillocks were 40− 80 nm high, which is slightly higher than the dendrite patterns 5894

dx.doi.org/10.1021/cg3012537 | Cryst. Growth Des. 2012, 12, 5889−5896

Crystal Growth & Design

Article

Furthermore, we showed that the previously proposed mechanism for the anticaking effectiveness for Fe-mTA12 is probably incorrect. A surface X-ray diffraction experiment is needed to determine exactly how this iron complex adsorbs on the {100} surface of sodium chloride crystals. Both anticaking agents influence the surface morphology of the sodium chloride crystals after separation of the crystals from the humid environment and subsequent drying. Ferrocyanide causes the formation of dendrites and cornered, square hillocks on the surface, whereas Fe-mTA causes the formation of almost hemispherical hillocks. This difference might explain the difference in flowability between ferrocyanide-treated and FemTA-treated sodium chloride powders.

The higher concentration is also caused by the fact that, during storage, the overall supersaturation (Δμ) is not constant but varies and the time available for caking is longer. The local supersaturation can be even higher near the contact points between adjacent NaCl grains, because of the reentrant contact angle and the local presence of more water due to the capillary effect. This locally creates an increase of the supersaturation and can lead to solid bridge formation between adjacent particles.34 For Fe-mTA, the concentration needed for the steps to be completely pinned corresponds well with the required amount for effective anticaking. Individual Fe-mTA complexes are not able to pin steps at very low concentrations, in contrast to ferrocyanide ions. At higher coverages (θ = 0.5), step pinning is observed for Fe-mTA. A possible explanation for this difference is that single Fe-mTA complexes are, in principle, able to pin steps, but most of the Fe-mTA is present as an inactive complex. This inactive part is, for instance, still dissolved in the adsorbed water layer because of a high desorption rate or immobilized as an inactive Fe-mTA complex. It is likely that not all of the Fe-mTA is effective because it is light-sensitive. Also, the complexation of Fe-mTA, and therefore its effectiveness, depends strongly on the pH and solvent composition. An alternative explanation is that single complexes of Fe-mTA are unable to pin steps and clusters of Fe-mTA complexes are needed. Both complexes adsorb onto the {100} surface of sodium chloride. Ferrocyanide ions can be adsorbed more strongly, because they are able to pin steps at very low concentrations. The interaction between the Fe-mTA complex and the sodium chloride crystal is probably weaker. For ferrocyanide, we recently determined how this iron complex adsorbs onto this surface using surface X-ray diffraction.9 However, for the FemTA complex, this is unknown. Because both complexes pin steps, but finally the steps pass the pinning point unhindered, we conclude that, for both anticaking agents, the complex is reversibly bound to the surface. When a complex desorbs, the step is of course no longer pinned. We also conclude that the previously proposed mechanism of growth inhibition by Fe-mTA,12 namely, the formation of an iron oxide layer, is unlikely. If such a layer were formed on the crystal, this would not be a reversible process because of the extremely low solubility of iron oxide. These layers would then need to be incorporated, or the crystals would not grow at all. We believe that the iron oxide layers are an artifact of the observation technique. Because LEIS is a vacuum technique, it is not possible to study the anticaking agents in contact with the sodium chloride crystal in situ, and the layers were likely formed when the adsorbed water layer was evaporated, depositing the dissolved anticaking agent on the surface, which then underwent oxidation.

■

■

ASSOCIATED CONTENT

S Supporting Information *

Several film files: two films of step propagation on clean sodium chloride crystal surfaces, one pretreated with ethanol and one with methanol; one film of step pinning by ferrocyanide at a coverage of θ = 0.0002; and one film of step pinning by FemTA at a coverage of θ = 0.5. Some image files of surface morphologies, such as those in Figure 8, at different magnifications. Text file explaining the choice of the experimental conditions. This material is available free of charge via the Internet at http://pubs.acs.org.

■

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS We acknowledge Akzo Nobel Industrial Chemicals and the Dutch Ministry of Economic Affairs for funding (EOS−KTO Program, AgentschapNL). Furthermore, we thank Jelle Eygensteijn for the ICP-MS measurements.

■

REFERENCES

(1) Nývlt, J.; Ulrich, J. Admixtures in Crystallization; VCH: Weinheim, Germany, 1995. (2) Sangwal, K. Additives and Crystallization Processes; Wiley: New York, 2007. (3) Kaufmann, D. Sodium Chloride: The Production and Properties of Salt and Brine; Reinhold Publishing Corporation: New York, 1960. (4) Chen, Y.; Chou, J. Powder Technol. 1993, 77, 1−6. (5) Cabrera, N.; Vermilyea, D. The Growth of Crystals from Solution. In Growth and Perfection of Crystals; Doremus, R. H., Roberts, B. W., Turnbull, D., Eds.; Wiley: New York, 1958; pp 393− 410. (6) Lacmann, R.; Stranski, I. The Effect of Adsorption of Impurtities on the Equilibrium and Growth Forms of Crystals. In Growth and Perfection of Crystals; Doremus, R. H., Roberts, B. W., Turnbull, D., Eds.; Wiley: New York, 1958; pp 427−439. (7) Weissbuch, I.; Addadi, L.; Lahav, M.; Leiserowitz, L. Science 1991, 253, 637−645. (8) Radenovic, N.; van Enckevort, W.; Kaminski, D.; Heijna, M.; Vlieg, E. Surf. Sci. 2005, 599, 196−206. (9) Bode, A.; Vonk, V.; Kok, D.; van den Bruele, F.; Kerkenaar, A.; Mantilla, M.; Jiang, S.; Meijer, J.; van Enckevort, W.; Vlieg, E. Cryst. Growth Des. 2012, 12, 1919−1924. (10) van Damme-van Weele, M. Influence of additives on the growth and dissolution of sodium chloride crystals. Ph.D. Thesis, Technische Hogeschool Twente, Twente, The Netherlands, 1965.

CONCLUSIONS

We were able to observe monatomic step movement on the {100} faces of sodium chloride crystals in contact with a humid atmosphere at 58% RH using AFM. In this way, we were able to determine the rate of step movement, the supersaturation, and the driving force. Also, we were able to examine how the anticaking agents ferrocyanide and Fe-mTA inhibit the growth of sodium chloride crystals by observing step pinning in situ. Ferrocyanide does this already at a concentration of θ = 2 × 10−4; for Fe-mTA, a coverage of θ = 1 is required. 5895

dx.doi.org/10.1021/cg3012537 | Cryst. Growth Des. 2012, 12, 5889−5896

Crystal Growth & Design

Article

(11) Geertman, R. Use of carbohydrate-based metal complexes in non-caking salt compositions. World Patent WO 00/59828, 2006. (12) Geertman, R. VDI-Ber. 2005, 1901, 557−562. (13) Teng, H.; Dove, P.; Orme, C.; Yoreo, J. D. Science 1998, 282, 724−727. (14) Land, T.; Martin, T.; Potapenko, S.; Palmore, G. T.; DeYoreo, J. Nature 1999, 399, 442−445. (15) Mauri, A.; Moret, M. J. Cryst. Growth 2000, 208, 599−614. (16) Orme, C.; Noy, A.; Wierzbicki, A.; McBride, M.; Grantham, M.; Teng, H.; Dove, P.; DeYoreo, J. Nature 2001, 411, 775−779. (17) Wilkins, S.; Coles, B.; Compton, R. J. Phys. Chem. B 2002, 106, 4763−4774. (18) Shindo, H.; Ohashi, M.; Baba, K.; Seo, A. Surf. Sci. 1996, 357− 358, 111−114. (19) Shindo, H.; Ohashi, M.; Tateishi, O.; Seo, A. J. Chem. Soc., Faraday Trans. 1997, 93, 1169−1174. (20) Peters, S.; Ewing, G. J. Phys. Chem. B 1997, 101, 10880−10886. (21) Foster, M.; Ewing, G. Surf. Sci. 1999, 427−428, 102−106. (22) Hallquist, M.; Petrucci, N.; Kreuzer, C.; Ostanin, V.; Cox, R. Phys. Chem. Chem. Phys. 2000, 2, 4373−4378. (23) Oshima, T.; Zhang, Y.; Hirota, M.; Suzuki, M.; Nakagawa, T. Adv. Powder Technol. 1995, 6, 35−45. (24) Jeong, H.; Williams, E. Surf. Sci. Rep. 1999, 34, 171−294. (25) Bartelt, N.; Tromp, R.; Williams, E. Phys. Rev. Lett. 1994, 73, 1656−1659. (26) Bartelt, N.; Theis, W.; Tromp, R. Phys. Rev. B 1996, 54, 11741− 11751. (27) Kellogg, G.; Bartelt, N. Surf. Sci. 2005, 577, 151−157. (28) Cleaver, J.; Wong, P. Surf. Interface Anal. 2004, 36, 1592−1599. (29) Cleaver, J.; Karatzas, G.; Louis, S.; Hayati, I. Powder Technol. 2004, 146, 93−101. (30) Ostwald, W. Lehrbuch der Allgemeinen Chemie; Englemann: Leipzig, Germany, 1896; Vol. 2. (31) Burton, W.; Cabrera, N.; Frank, F. Proc. R. Soc. A 1950, 243, 299−358. (32) Christoffersen, J.; Rostrup, E.; Christoffersen, M. J. Cryst. Growth 1991, 113, 599−605. (33) van Enckevort, W.; Los, J. J. Phys. Chem. C 2008, 112, 6380− 6389. (34) Wahl, M.; Kirsch, R.; Trapp, S.; Bottlinger, M. Chem. Eng. Technol. 2006, 29, 674−678.

5896

dx.doi.org/10.1021/cg3012537 | Cryst. Growth Des. 2012, 12, 5889−5896